CN106054611A - ECG baseline drift suppression method based on model-free adaptive prediction control - Google Patents

Abstract

The invention discloses an ECG baseline drift suppression method based on model-free adaptive prediction control. The method includes that expected output (as shown in the description) is preset, the data being free of baseline drift is selected as the expected output, and the baseline value is 0. An improved model-free adaptive prediction control algorithm is used to calculate the input control signal Delta u(k), and then a pseudo partial derivative (as shown in the description) is subjected to prediction estimation and identification, and at the end, an A1 matrix is constructed based on the pseudo partial derivative, and furthermore, a control system is constructed. The control signals act on the constructed system to achieve the effect of ECG baseline drift suppression. According to the invention, the ECG baseline drift suppression during the ECG acquisition is a very meaningful innovation method which is high in versatility, fast in speed, and good in effect.

Description

Electrocardio baseline drift suppressing method based on model-free adaption PREDICTIVE CONTROL

Technical field

The present invention relates to medical instruments field, after a kind of improvement, calculate based on model-free adaption PREDICTIVE CONTROL The electrocardio baseline drift suppressing method of method.

Background technology

Electrocardiogram (electrocardiogram, ECG) is the potential change figure that cardiomotility is relevant, and it is diagnosis A kind of modern technologies of disease, especially in terms of making a definite diagnosis and differentiating various arrhythmia, relatively other diagnose electrocardiographic diagnosis method Method is the most reliable.Owing to ECG is the same with most of biomedicine signals, it is all the small-signal that signal-noise ratio is the lowest, often Being mixed with the strongest background noise, main throat sound is Hz noise, myoelectricity interference, baseline drift etc., to letter in various noises What number impact was maximum is baseline drift throat sound.Therefore, during ECG detection identifies, first eliminate baseline drift.Baseline drift Because breathing, caused by limb activity or exercise electrocardiogram test, therefore so make the datum line of ECG signal present and float up and down Situation about moving.It is a kind of slowly varying low frequency signal.In order to eliminate the interference of baseline drift, have pointed out both at home and abroad and apply Many methods.Filter method and matching base drift method are to remove two class main method of electrocardio baseline drift.Include medium filtering, FIR filtering, wavelet transformation etc. remove the filtering method of base drift.These methods applying at removal electrocardio baseline the most many times In the experiment of drift:

(a) medium filtering: medium filtering is removed the general flow of base drift and is: ECG signal S containing base drift by window width is The median filter of 200ms, obtains eliminating the signal S1 of QRS wave and P ripple.S1 is the median filter of 600ms by window width, Then T ripple is the most disallowable, finally gives base drift interference S2.

B () FIR filters: N rank FIR filter has N number of zero point in Z plane, and its convergence domain always includes unit circle, therefore FIR filter is the most stable.The FIR filter of coefficient symmetry can keep linear phase, this be ECG remove make an uproar required, But should be noted that N rank even symmetry or odd symmetric FIR filter have N/2 point time delay.The principle of FIR filtering is simple, without cumulative error, Be suitable to real-time occasion, but amplitude-frequency distortion is the most obvious.

(c) wavelet transformation: the wavelet transformation of signal is equivalent to signal and leads to and low-pass filtering, decomposable asymmetric choice net at the band of different scale Obtain approximation component and the details coefficients of signal.Baseline drift is mainly low frequency component, during wavelet reconstruction, as long as will The approximation component zero setting of high yardstick, has just obtained removing the signal of base drift.In the case of signal sampling frequency is certain, for certain One determines small echo, and center frequency under its different scale and window width all determine that, thus can determine that and removes the little of baseline drift The optimal decomposition scale of wave conversion.

These methods are respectively arranged with advantage, but are all that the optimization carrying out the electrocardiogram (ECG) data obtained processes, although be respectively arranged with advantage, but Also there is the restriction of correspondence.

Summary of the invention

It is an object of the invention to overcome the existing methodical shortcoming and defect being mentioned above, it is proposed that a kind of based on nothing The electrocardio baseline drift suppressing method of model adaptation PREDICTIVE CONTROL, during collection is cardiac electrical, the just removal of pretreatment Falling cardiac electrical baseline noise, it is intended to wearable for develop in the future, ambulatory medical device etc. provides good theory and application base Plinth.

For achieving the above object, the technical solution adopted in the present invention is:

A kind of electrocardio baseline drift suppressing method based on model-free adaption PREDICTIVE CONTROL, it is characterised in that include step Rapid:

A () arranges desired output

B () calculates according to MFA control algorithm controls criterion and controls input signal Δ u (k)；

(c) pseudo-partial derivativePredicted estimate and identification；

D () builds A1 matrix according to pseudo-partial derivative, and then build control system.

Further, step (a) select the data without baseline drift as desired output, it is simply that baseline value is 0.

Further, the forecast model in step (b) is:

WhereinΔ u (k)=u (k)-u (k-1)；

Being the desired output in K+1 moment, y (k) is the actual output in system k moment, and u (k) is the reality in k moment Input, u (k-1) is actually entering of k-1 moment, and the purpose of control is to apply to control u (k) to system in the k moment, it is desirable to k+1 The output in moment is desired for

Further, step (b) introduces constant, τ time lag, between two groups of input values of band constant, τ time lag Rate of change, as input criterion function in an important constraint, formula is:

$\mathrm{\Δ}u\left(k\right)={\[\frac{\left(u\right(k)-u(k-1-\mathrm{\τ}\left)\right)}{T}\]}^2.$

Further, step (c) uses the input criterion function after improving, and the pseudo-partial derivative after improvement is:

Wherein, μ>0 is weight factor, and 0<η≤1 is step factor, then according to Forecasting Methodology, carries out pre-to pseudo-partial derivative Survey.

Further, in step (d), A1 comprises in systemPass through structure The A1 matrix built, builds control system.

Accompanying drawing explanation

Fig. 1 is Electrocardiograph control figure；

Fig. 2 is method implementing procedure figure；

Fig. 3 is electrocardio baseline and the original electrocardiographicdigital baseline of prediction and controls input signal；

Fig. 4 is the electrocardiosignal after original electro-cardiologic signals and suppression baseline drift；

Fig. 5 is the predictive value of pseudo-partial derivative.

Detailed description of the invention

Below in conjunction with embodiment and accompanying drawing, the present invention is described in further detail, but embodiments of the present invention do not limit In this.

Fig. 1 is Electrocardiograph control figure, exports control signal, effect by the parameter of model-free adaption system MFAC algorithm Electrocardiograph, reaches the effect of electrocardio baseline, the actual output of return and desired output difference, carries out feedback regulation, protects Card electrocardio goes the accuracy of baseline.

A kind of electrocardio baseline drift suppressing method based on model-free adaption forecast Control Algorithm, is embodied as flow process such as Shown in Fig. 2, it is achieved step is as follows:

A () arranges desired output

As desired output, being because preferably removing baseline effects is exactly that baseline drift does not all have.

B () calculates according to MFA control algorithm controls criterion and controls input signal Δ u (k):

General forecast model:

Wherein

Being the desired output in K+1 moment, y (k) is the actual output in system k moment, and u (k) is the reality in k moment Input, u (k-1) is actually entering of k-1 moment, and the purpose of control is to apply to control u (k) to system in the k moment, it is desirable to k+1 The output in moment is desired for

Herein, the present invention be introduced into one time lag constant, τ, between two groups of input values of band constant, τ time lag Rate of change, as input criterion function in an important constraint.On the basis of former method, increase stablizing of system Property, the anti-interference and tracing property of the system of lifting, reach the accuracy of denoising.As follows:

$\mathrm{\&Delta;}u\left(k\right)={\&lsqb;\frac{\left(u\right(k)-u(k-1-\mathrm{\&tau;}\left)\right)}{T}\&rsqb;}^2$

(c) pseudo-partial derivativePredicted estimate and identification:

Using the input criterion function after improving, the pseudo-partial derivative after improvement is:

Wherein, μ>0 is weight factor, and 0<η≤1 is step factor.Then according to Forecasting Methodology, pseudo-partial derivative is carried out pre- Survey.

D () A1 comprises in systemBy the A1 matrix built, build Control system.

By the system built, use the effect of input control signal, obtain the output of reality.It is right that the effect of regulation is The regulation of input control signal and the estimation of pseudo-partial derivative.By actual output and the difference effect of desired output, difference is anti- Feedback acts on input control signal, and adaptive regulation reaches optimal output effect.Wherein, it was predicted that electrocardio baseline and original Electrocardio baseline and control input signal as it is shown on figure 3, electrocardiosignal such as Fig. 4 after original electro-cardiologic signals and suppression baseline drift Shown in, the predictive value of pseudo-partial derivative is as shown in Figure 5.

Above-described embodiment is the present invention preferably embodiment, but embodiments of the present invention are not by above-described embodiment Limit, the change made under other any spirit without departing from the present invention and principle, modify, substitute, combine, simplify, All should be the substitute mode of equivalence, within being included in protection scope of the present invention.

Claims (6)

1. an electrocardio baseline drift suppressing method based on model-free adaption PREDICTIVE CONTROL, it is characterised in that include step:

A () arranges desired output

B () calculates according to MFA control algorithm controls criterion and controls input signal Δ u (k)；

(c) pseudo-partial derivativePredicted estimate and identification；

D () builds A1 matrix according to pseudo-partial derivative, and then build control system.

Method the most according to claim 1, it is characterised in that: step (a) select the data without baseline drift as the phase Hope output, it is simply that baseline value is 0.

Method the most according to claim 1, it is characterised in that: the forecast model in step (b) is:

WhereinΔ u (k)=u (k)-u (k-1)；

Being the desired output in K+1 moment, y (k) is the actual output in system k moment, and u (k) is that the reality in k moment is defeated Entering, u (k-1) is actually entering of k-1 moment, and the purpose of control is to apply to control u (k) to system in the k moment, it is desirable to during k+1 The output carved is desired for

Method the most according to claim 1, it is characterised in that: step (b) introduces constant, τ time lag, by the band time Rate of change between two groups of input values of hysteresis constant τ, as an important constraint in input criterion function, formula is:

$\mathrm{\&Delta;}u\left(k\right)={\&lsqb;\frac{\left(u\right(k)-u(k-1-\mathrm{\&tau;}\left)\right)}{T}\&rsqb;}^2.$

Method the most according to claim 1, it is characterised in that: step (c) uses the input criterion function after improving, and improves After pseudo-partial derivative be:

Wherein, μ>0 is weight factor, and 0<η≤1 is step factor.Then according to Forecasting Methodology, pseudo-partial derivative is predicted.

Method the most according to claim 1, it is characterised in that: in step (d), A1 comprises in systemBy the A1 matrix built, build control system.